OmniPanda is built on AI. This page explains, without marketing language, what the AI actually does, where it fails, what it is allowed to do with your money, and who is responsible for the result. It is part of our Terms of Service.
If you read only one paragraph, read this one: OmniPanda's AI can be wrong. It is not a licensed adviser, it is not a guarantee of anything, and in Automatic mode it can place real trades in your real brokerage account without asking you first. You choose whether to turn that on, and you remain responsible for your account.
1. Where AI is used
AI is involved in most of what OmniPanda does:
1.1 The chat assistant answers your questions, researches companies and markets, reads your connected accounts, explains your portfolio, and decides which of its tools to call to get an answer.
1.2 Research and summaries — company overviews, news summaries, bull and bear cases, comparisons, and the written parts of reports.
1.3 AI Portfolio — building a portfolio approach from your investor profile, reviewing an enrolled account on a schedule, deciding what to buy and sell, and writing the reasoning you see in a trade plan.
1.4 Trade preparation — turning "buy some of this" into a concrete order for you to review in the trade widget.
1.5 Voice mode — turning your speech into text, and the reply back into speech.
1.6 Strategies and backtests — turning a plain-English description of a strategy into rules, and explaining the results.
1.7 Small helpers throughout the app — titles, labels, categorisation, search relevance and similar.
2. How it works, in plain terms
2.1 OmniPanda sends your request, along with the context it needs, to AI models — including large language models — run by AI providers we contract with. We do not build the underlying models ourselves.
2.2 The context we send can include: your message, your recent conversation, the page you are looking at, relevant parts of your investor profile, and relevant account data such as your positions, balances and recent orders. The model uses tools we provide to fetch market data, read your accounts, run calculations, and — where you have authorized it — place orders.
2.3 The model produces text, structured output such as a trade plan, or a tool call. Nobody at OmniPanda reads or approves that output before you see it. It is generated and delivered automatically, in seconds.
2.4 Models change. We update, replace and tune the models and prompts behind the Service, and different features may use different models. The same question can get different answers on different days, or even twice in a row.
2.5 AI output is not a calculation from a verified source of truth. It is a prediction of a plausible answer. It is often right. It is not reliably right.
3. What the AI gets wrong
These are not hypothetical. They are the normal failure modes of this technology, and you should expect to meet them:
3.1 It invents things. A model can state a fact, a figure, a date, a quote or a source that does not exist, in a confident and well-written sentence. This is called hallucination, and no current model is free of it.
3.2 It uses stale data. A model's own knowledge has a cut-off date. Prices, filings, news, analyst views and even a company's name or listing status can have changed since. Data we fetch for it can also be delayed or out of date.
3.3 It makes arithmetic and unit mistakes — a percentage confused with a fraction, millions with billions, a currency mixed up, a return calculated on the wrong base, a total that does not add up.
3.4 It misreads your situation. It sees what we send it. It does not know about your other accounts, your job, your debts, your family, your taxes, your other plans, or anything you have not told it.
3.5 It is inconsistent. It can contradict itself in the same conversation, or reverse a view without new information.
3.6 It sounds certain when it is not. Fluency is not accuracy. A polished answer and a wrong answer look the same.
3.7 It has blind spots and biases inherited from its training data — towards large, well-covered US companies, towards recent narratives, and towards whatever was written most often.
3.8 It can be steered. Content it reads — a news article, a filing, a document — can contain instructions or misinformation that influence its answer.
3.9 It can fail quietly. A tool call can return incomplete data, or nothing, and the model may still produce a confident-sounding answer around the gap.
3.10 Nothing is verified, and nothing is updated. AI output is not checked by a person for factual accuracy before you see it, and once it is written it is never revised. An answer you got last week reflects what the model produced last week, and we have no obligation to go back and correct it.
4. AI output is not investment advice
4.1 Nothing OmniPanda's AI produces is investment advice, a recommendation that an investment or strategy is suitable for you, a solicitation, an offer, a research report in the regulatory sense, tax advice, legal advice or accounting advice. Pandamensional Inc. is not a registered investment adviser and is not a broker-dealer, and we do not give personalised investment advice. What the AI produces is information and automation that you direct. In Automatic mode it places trades in an account you enrolled, within the limits you chose.
4.2 The AI does not know whether something is right for you. It has no duty of loyalty to you, no duty of care, and no professional licence. It is a tool.
4.3 A trade plan is a proposal produced by software from the profile and limits you gave it — not a professional's opinion about your finances.
4.4 Projections, targets, scores, probabilities and backtests are model output about an uncertain future. They are not forecasts you can rely on, and past performance — real or simulated — does not predict future results.
4.5 Naming a security is not recommending it. The assistant talks about specific companies, funds and tickers, scores them, compares them and puts them in plans. None of that is us telling you that buying, selling or holding one is right for you.
4.6 Explanations are patterns, not causes. When OmniPanda says your portfolio moved because of something, or that a factor is driving a stock, it is describing a statistical or narrative association it found in data — not a proven cause, and not a mechanism it has verified.
4.7 For advice about your own money, talk to a licensed financial adviser, an accountant or a lawyer.
5. When the AI can act on its own: Automatic mode
This is the part of OmniPanda with the most consequence, so here it is in full.
5.1 What it is. When you enroll a connected brokerage account in AI Portfolio, you choose one of two modes. In Approval Required, every review produces a trade plan and nothing is placed until you approve it. In Automatic mode, the AI places the trades itself as soon as a review finishes — no per-trade approval, no per-plan approval, and no second question.
5.2 You turn it on. OmniPanda never enrolls an account for you and never moves an account into Automatic mode for you. Automatic mode applies only to the accounts you enrolled yourself, only at brokerages that support trading through OmniPanda, and only while you leave the account in that mode. Enrollment starts in Approval Required, and choosing Automatic — while enrolling, or later from the account's settings — asks you to confirm that OmniPanda may trade without asking you each time. We ask on every such switch you make in the app. The enrollment screen also shows the mode and a summary before you confirm — read it before you finish.
5.3 The limits are yours. Before it manages an account, you give it an investor profile and set the guardrails: how often it reviews the account, how much of the portfolio may change hands in a single review, how many orders a single review may place, and how concentrated a single position or a single sector may become. The AI is designed to work inside those limits and to skip or hold a trade that would break one.
5.4 The limits are not a safety net. They are constraints on the AI's behaviour, not protection for your money. A portfolio that stays entirely inside your limits can still lose a lot. And because this is software, we cannot promise a limit is honoured in every case — a bug, bad data or unexpected brokerage behaviour can defeat one.
5.5 What the AI can do. Read the enrolled account, decide on trades, and place, and where possible cancel, orders in that account through your brokerage connection.
5.6 What the AI cannot do. It cannot withdraw cash, transfer securities, move money to another person or account, change your brokerage's settings, open or close a brokerage account, or touch an account you did not connect and enroll. We are not a broker and we never hold your money or your investments — they stay at your own brokerage the whole time.
5.7 You can stop it at any time. Switch the account back to Approval Required, pause or stop the portfolio, or disconnect the brokerage. Any of those stops future automatic trading. Stopping does not undo trades already placed and does not cancel an order already sent to your brokerage.
5.8 It is not watching your positions. Reviews happen on the schedule you chose — daily at the most. Between reviews, nothing is monitoring your account. Automatic mode is not a stop-loss, not a hedge, not a circuit breaker and not downside protection. A market can move a long way between two reviews.
5.9 Timing is not guaranteed. A review can run late, run at an unexpected hour, or not run at all — because of an outage at us, at your brokerage or at a data provider, or because the market was closed. Orders can fill at a worse price, fill partly, fill late, or be rejected.
5.10 It can be wrong with real money. Every failure mode in section 3 can express itself as a trade. Automation makes a mistake faster and, if the mistake is in the model's reasoning, it can repeat it.
5.11 We may pause it. We may suspend automatic trading for an account or for everyone — during an outage, a suspected bug, a data problem, a security concern or a billing problem. We will try to tell you, but we may have to act first.
5.12 You give up your brokerage's own checks. Placing an order yourself means seeing your brokerage's confirmation screen, its warnings, its fee and margin disclosures, and having a last chance to stop. An order OmniPanda places skips all of that. Your brokerage treats the instruction as coming from you and applies its own rules — margin, pattern-day-trading, permissions — as though you had typed it in. If your account allows margin, a trade may use it.
5.13 Every run is on the record. Each review, the plan it produced, the trades it placed or skipped, and the AI's reasoning are saved in your run history. If something looks wrong, that is where to look first — and then tell us.
6. What you are responsible for
6.1 The account is yours. The positions, the gains, the losses, the fees, the interest, the dividends and the taxes are yours — including for trades the AI placed under the authority you gave it.
6.2 Check anything that matters. Before acting on what the AI tells you, verify it: prices and quotes at your brokerage or an exchange, holdings and balances at your brokerage, order details on the confirmation screen, company facts in the company's own filings, and tax treatment with a tax professional. Your brokerage's records are the authoritative ones, not ours.
6.3 Do not rely on the AI alone, and do not treat a fluent answer as a verified one.
6.4 Do not rely on OmniPanda being available. If you have an enrolled account, have a plan for what you would do if OmniPanda stopped working, or stopped existing, tomorrow.
6.5 Stay in Approval Required if you are new to this — it is where every enrollment starts. Watch a few plans, see whether you agree with the reasoning, and only then decide whether to go hands-off.
7. What we do to reduce the risk
We are not going to claim the AI is safe. Here is what we actually do:
7.1 Hard limits on each enrolled account — turnover, order count, and position and sector concentration — checked before an order is placed, outside the model's discretion.
7.2 Trades are only possible in accounts you explicitly connected and enrolled, at brokerages that support trading, and only through your own brokerage connection.
7.3 The AI has no ability to move money or securities out of your account. There is nothing in the design for it to abuse.
7.4 Approval Required mode exists, and is there for exactly this reason.
7.5 Every run, plan and piece of reasoning is logged and shown to you, so a bad decision can be found rather than hidden.
7.6 Market and account data is fetched from your brokerage and from data providers rather than recalled from the model's memory, so the model reasons over retrieved data — though that data can still be delayed or wrong.
7.7 We monitor errors and can pause features, or automatic trading, when something looks wrong.
7.8 None of this makes the AI correct. It makes its mistakes smaller, slower and visible.
8. Your data and AI
8.1 What is sent. To answer you or manage an enrolled account, we send AI providers the context described in section 2.2 — your message or the task, relevant conversation history, relevant profile information, and relevant account data such as positions and balances. We do not send your password or your card details.
8.2 Voice. In voice mode your audio is sent to be transcribed, and replies are sent to be turned into speech. We do not keep your audio after it has been transcribed. The transcript is kept as part of your chat history.
8.3 Training. We do not allow the AI providers we use to train their general-purpose models on your content, and we do not sell your content to anyone for that purpose. We do review examples of AI output — and, where we need to, the conversation around it — to find and fix bugs and bad answers. We use aggregated or de-identified information to improve our prompts, our tools and our own testing.
8.4 Storage. Your chats, including tool results and the plans and reasoning from portfolio runs, are stored in your account so you can come back to them and so a later conversation has context. You can ask us to delete your account and data at any time; our Privacy Policy explains how, and what we have to keep.
8.5 Full details of what we collect, why, who we share it with and how long we keep it are in our Privacy Policy. It is linked at the bottom of this page.
9. Telling us something went wrong
9.1 If the AI told you something false, produced a plan that does not make sense, placed a trade you did not expect, or broke one of your limits, tell us at info@pandamensional.ai.
9.2 Please include: your account email, the account or portfolio involved, roughly when it happened, the run or conversation if you can point us at it, and what you expected instead. If it was a trade, include the symbol and the date.
9.3 If money is moving right now and you need it to stop, do not wait for us. Switch the account to Approval Required, stop the portfolio, or disconnect the brokerage — and contact your brokerage directly about an order already placed. Your brokerage, not OmniPanda, controls your open orders and your account.
9.4 We read every report. We look for a bug, a data problem or a prompt problem behind it, and we fix what we can. Reporting a problem does not by itself mean we are liable for it; the Terms of Service govern that.
9.5 You can also just ask for a human. Email us and a person will answer — typically within one business day.
10. Changes to this disclosure
We will update this page as the Service and the models behind it change, and we will change the "Last updated" date when we do. If a change materially affects what the AI is allowed to do with your accounts, we will tell you by email or in the app before it takes effect.
11. Contact
- Pandamensional Inc.
- 103 Mint Pl, Athens, GA 30606, USA
- info@pandamensional.ai